Apply machine learning techniques to detect malicious network traffic in cloud computing

• Main Authors: Amirah Alshammari, Abdulaziz Aldribi
• Format: Article
• Language: English
• Published: SpringerOpen 2021-06-01
• Series: Journal of Big Data
• Subjects: IDS; Network traffic; Feature extraction; Dataset; Machine learning
• Online Access: https://doi.org/10.1186/s40537-021-00475-1

Abstract Computer networks target several kinds of attacks every hour and day; they evolved to make significant risks. They pass new attacks and trends; these attacks target every open port available on the network. Several tools are designed for this purpose, such as mapping networks and vulnerabilities scanning. Recently, machine learning (ML) is a widespread technique offered to feed the Intrusion Detection System (IDS) to detect malicious network traffic. The core of ML models’ detection efficiency relies on the dataset’s quality to train the model. This research proposes a detection framework with an ML model for feeding IDS to detect network traffic anomalies. This detection model uses a dataset constructed from malicious and normal traffic. This research’s significant challenges are the extracted features used to train the ML model about various attacks to distinguish whether it is an anomaly or regular traffic. The dataset ISOT-CID network traffic part uses for the training ML model. We added some significant column features, and we approved that feature supports the ML model in the training phase. The ISOT-CID dataset traffic part contains two types of features, the first extracted from network traffic flow, and the others computed in specific interval time. We also presented a novel column feature added to the dataset and approved that it increases the detection quality. This feature is depending on the rambling packet payload length in the traffic flow. Our presented results and experiment produced by this research are significant and encourage other researchers and us to expand the work as future work.